22 January 1998
Physics Letters B 417 Ž1998. 297–302
Systematic proof of the existence of Yangian symmetry in chiral Gross–Neveu models Tamas Hauer
1
Institute for Theoretical Physics, Lorand EotÕos ¨ ¨ UniÕersity, H-1088 Budapest, Puskin u. 5-7, Hungary Received 4 September 1997; revised 11 October 1997 Editor: L. Alvarez-Gaume´
Abstract The existence of non-local charges, generating a Yangian symmetry is discussed in generalized chiral Gross-Neveu models. Their conservation can be proven by a finite-loop perturbative computation, the order of which is determined from group theoretic constants and is independent of the number of flavors. Examples, where the 1-loop calculation is sufficient, include the SO Ž n.-models and other more exotic groups and representations. q 1998 Elsevier Science B.V. PACS: 11.10.Kk; 11.20.Ex; 11.30.Na Keywords: Non-local charge; Operator product expansion; Gross-Neveu model; Renormalization group
1. Introduction In the theory of two dimensional integrable systems a fundamental role is played by those conserved quantities which guarantee the solvability of the models. While in many cases classical integrability is manifest, the proof of conservation and even the proper definition of these charges may be quite subtle in the corresponding quantum theory. In certain models where the symmetry algebra is generated by nonlocal charges w1,2x the approach originally w3x proved to be fruitful. In proposed by Luscher ¨ these Žasymptotically free. field theories, the properties of nonlocal expressions of the local currents can be traced back to the short distance singularities of 1 E-mail address:
[email protected]. Address: Center for Theoretical Physics, MIT, Cambridge, MA 02139.
the current algebra. The existence of Luscher’s non¨ local conserved charge is the consequence of the fact that the operator product expansion ŽOPE. of the currents close on themselves and their derivatives, which replaces, in the quantum theory, the zerocurvature condition of the classical currents. ŽSee w4x for a summary and w5x for a complete review.. The question about the closure of the current algebra is a delicate one, and the answer varies from one model to another. In the simplest case Žlike the O Ž n. nonlinear sigma model w3x, a large class of generalized sigma models w6x or the chiral SUŽ n. Gross-Neveu ŽGN.-model w5x. there are too few degrees of freedom to form operators which may ruin the conservation, while in other theories, extra fields may be constructed and it is the dynamics of the model which determines their ultimate presence or absence. For example, in the CP Ny 1-model, the conservation
0370-2693r98r$19.00 q 1998 Elsevier Science B.V. All rights reserved. PII S 0 3 7 0 - 2 6 9 3 Ž 9 7 . 0 1 4 3 1 - 7
298
T. Hauerr Physics Letters B 417 (1998) 297–302
is ruined by the extra term w7x, while in its supersymmetric partner this quantum anomaly disappears w8x. In a previous paper w4x we studied the SUŽ n. Žmultiflavor., chiral Gross-Neveu model ŽŽM.CGN. w9x, where - thanks to renormalization group invariance a one-loop perturbative computation proved to be decisive and saved the desired form of the current algebra. The aim of the present letter is to extend this argument to a general class of chiral GN-type models, which are defined as Lagrangian field theories with current-current interaction, and where the symmetry algebra generated by the currents is an arbitrary simple Lie-algebra. This family of theories Žwithout flavor multiplicity. was studied in w10x where, the non-local charges generating the classical Yangian were constructed. Our goal is to investigate whether the conservation of the non-local generator of the algebra survives the quantization using the above strategy. We will explicitly calculate the leading exponent of the perturbative coupling constant of the OPE-coefficients in terms of group-theoretic constants and obtain a simple formula ŽEq. Ž14.., expressing the order of the needed perturbative calculation in terms of quadratic Casimirs of certain representations. We also give a large class of examples where zero- or one-loop results yield conclusion and show that the multiplicity Žflavor. of the fermion field does not affect the question. These theories include the SO Ž n. models 2 and a bunch of previously uninvestigated 3 models characterized by different groups, representations and flavor multiplicity. The plan of the paper is as follows. First, we summarize the important points in the connection between the non-local conservation laws and the OPE of the current, then shortly review the RG-argument developed in w4x showing how the anomalous dimension of certain operators play fundamental role in the analysis. When generalizing to arbitrary groups, we then express these quantities in terms of quadratic Casimirs of certain representations, deriving the formula 14, which determines the order of
2 For the Žone-flavor. O Ž n.-symmetric GN-models, non-local Ward identities were proven in leading order of the 1r n-expansion in w11x 3 This refers to the quantum case, the study of the classical conservation laws is similar for the whole family w10x
the needed perturbative calculation. Finally, we present examples where no more then a 1-loop computation is sufficient to reach conclusion.
2. OPE and non-local charge In the field theories under consideration we have a set of conserved, local currents: Em j a m Ž x . s 0, transforming under the adjoint representation of a simple Lie-group, G , with charges satisfying 4 : Q a ,Q b s f a b c Q c .
Ž 1.
In addition we require the QFT be renormalizable and asymptotically free and that all operators in the adjoint representation of G have higher canonical dimension than the current. If these conditions hold then, the general form of the current-current OPE Žup to vanishing terms as x ™ 0. is the following: r f a b c jmb Ž x . jnc Ž 0 . sCmn Ž x . jra Ž 0 . sr q Dmn Ž x . Es jra Ž 0 . q Ý Ei Oiwa mn x Ž 0 . ,
Ž 2.
i r Ž . sr Ž . where Cmn x and Dmn x are functions with leading singularities O Ž< x
Q1a s
1
`
H dy dy 4 y` 1
`
2e
Ž y1 y y 2 . f a b c j0b Ž t , y1 . j0c Ž t , y 2 .
a 1
Hy`dyj Ž t , y . ,
q
Ž 3.
can always be consistently defined, but it is conserved only if all the Ei ’s are zero Žboth statements are independent of the concrete form of j Ž x 2 ... Phrasing the condition in the above form immediately yields the straightforward strategy for proving the existence of the quantum charge: one assumes the presence of every antisymmetric tensor operator
4
Q a 'Hdxj a0 Ž x,t .; f a b c f b c d syC a d j d a d .
T. Hauerr Physics Letters B 417 (1998) 297–302
allowed by the symmetries, and computes the corresponding OPE-coefficients in some way; if all of them are zero then the conservation is proved. This program was successfully carried out in the CP Ny 1model with fermions, where supersymmetry prevented the classical value from receiving quantum corrections w8x; in our case renormalization group invariance will be of great help.
3. OPE and renormalization group As we showed in w4x, the calculation of the OPEcoefficients can be done using perturbation theory in models without dimensionful parameters, since in this case renormalization group invariance highly restricts the possible form of Ei . In the SUŽ n. MCGN we faced one extra operator and we expressed the leading exponent of its perturbative expansion in terms of its 1-loop anomalous dimension, however – as we will see – the argument is neither specific to SUŽ n. nor is it restricted to the extra operator standing alone. For simplicity, assume that the extra operators under consideration renormalize multiplicatively Ždo not mix with others. and the anomalous dimension of Oiwa mn x is given by D log Ž Zi . s hi ,1 g 2 q hi ,2 g 4 q . . . ,
Ž 4.
where the lhs. is the renormalization-group derivative of the renormalization constant Zi , corresponding to Oiwa mn x and g is the perturbative coupling of the model. For the OPE-coefficient, Ei s Ei ,0 g 2 a i Ž 1 q Ei ,1 g 2 q Ei ,2 g 4 q . . . . ,
Ž 5.
the renormalization group equation yields the following relation between a i , hi,1 , and the one-loop beta function coefficient, b 0 : hi ,1 ai s y . Ž 6. 2 b0 This equation is the key point in the argument since a i is either a positive integer, in which case it determines the order of the perturbative calculation needed to decide whether Ei vanishes or not, or if it is negative or non-integer then it does not allow Oiwa mn x to be present in the OPE.
299
4. Chiral Gross-Neveu models Now we turn to the CGN models, they are defined by the following Minkowskian action:
ž
S s d 2 x c i Euc y
H
g2 2
/
jma j a m .
Ž 7.
The fermionic field, c transforms under the irreducible representation, R Žwhereas c transforms under R . of the simple Žcolor. Lie-group, G and in case of the multiflavor models it also carries a multiplicity Žflavor. index. The current in the interaction Lagrangian is defined as jma Ž x . ' c Ž x . T agm c Ž x . ,
Ž 8.
where T a are generator matrices in representation R Žin case of the multiflavor models summation over the multiplicity indices is understood.. In order not to get in conflict with our assumption that the color current is the operator in the adjoint representation of G of the lowest canonical dimension, we require that the decomposition of R m R into irreducible representations does not contain the adjoint more than once. The first steps in determining the operator content of the OPE 2 are: collecting the set of operators allowed by symmetry and canonical dimensional analysis; and then – following our method – calculating their 1-loop renormalization. Apart from the current, there is one bilinear operator in the adjoint representation, iŽ Em c T agn c y c T agn Em c . which, however, has opposite C-parity to the current’s and is excluded. Therefore canonical dimensional analysis allows only for operators which are quadrilinear expressions of the fermionic field. In order to identify these fields we search for operators in the adjoint representation composed of the direct product R m R m R m R that is, we have to decompose this quadratic product into irreducibles. The following sequence will prove to be useful: decompose first R m R Žform bilinears. and then the pairwise products Žform bilinear of bilinears. from the two direct sums and look for the adjoint representations:
Ž R mR. m Ž RmR. s Ž . . . [ R1 [ . . . . m Ž . . . [ R 2 [ . . . . s . . . [ R ad j [ . . .
Ž 9.
T. Hauerr Physics Letters B 417 (1998) 297–302
300
It turns out that this decomposition will guarantee the multiplicative renormalization and it is the Casimir of R 1 and R 2 ‘‘defined’’ above that enters the anomalous dimension. To write down explicitly the operator we have in mind, denote the projectors on the basis elements of R 1 and R 2 with C 1 a and C 2 r , respectively and the one on the basis elements a in the adjoint representation by har : O wamn x ' haa r cg w m C 1 ac . Ž cgn x C 2 rc .
ž
/
Ž 10 .
To calculate the anomalous dimension of this operator one has to compute the one-loop four-particle correlation function. This contains four divergent Feynman-diagrams the sum of which is proportional to the following expression:
r s haa r Ž Ž C 1 a T b . i j Ž C 2 r T b . k l yŽ C1 a T b . i j Ž T b C 2 r . k l qŽ T b C1 a . i j Ž T b C 2 r . k l yŽ T b C1 a . i j Ž C 2 r T b . k l . ,
Ž 11 .
where i..l are color indices. If one treats C 1 and C 2 as tensor operators acting on R and recalls their commutation relation with T b this can be rewritten in terms of the generators t 1 b and t 2 b on representations R 1 and R 2 , respectively: a r s haa r t 1 b a X
Ž
1 sy 2
r 2b rX
X
1
5. Examples
X
. Ž t . Ž C1a . i j Ž C 2 r . kl
Ž CR
which determines the order at which the needed perturbative calculation becomes ‘‘exact’’ - may be such a small integer that this calculation can be done in finite amount of time Žas in w4x, where it was 1. or hopefully non-integer, in which case simply no real computation is needed. Notice furthermore that, since 11 contains the flavor indices in a trivial way, the multiplicative renormalization is also true for operators with nontrivial flavor-structure Žin the multiflavor models. and the same formula applies to their anomalous dimension. Let us summarize the above in the following recipe. Take a CGN-model, which is defined by G , R and the number of flavors. Decompose the product of representations Ž R m R . m Ž R m R . into irreducibles and find the adjoints; following the proposed decomposition, to every copy of the adjoint representation correspond two other ones, R 1 and R 2 . Calculate a i using 14 for every case obtained and take the highest nonnegative integer, a out of them. Calculate the OPE 2 up to a loops perturbatively and see whether it closes on the color currents or not; the fact you obtain is exact.
q CR 2 y C a d j . haa r Ž C 1 a . i j Ž C 2 r . k l ,
Ž 12 . where, we also used the tensor transformation properties of h a. Thus the operator renormalizes multiplicatively at one-loop order and its anomalous dimension is determined by the quadratic Casimirs CR 1, CR 2 , C a d j , of representations R 1 , R 2 and the adjoint: 1 h1 s Ž 13 . Ž C q CR 2 y C a d j . , 2p R 1 and this together with the one-loop b-function, b 0 s y C4apd j yields our magic number, a : CR q CR 2 y C a d j as 1 Ž 14 . Ca d j Eq. 14 is the main technical result of this paper. The power of this simple formula resides in that, a -
We now consider applications of 14 to various CGN models and look for the ones where no real computation is needed. In Ref. w4x we calculated the OPE-coefficients up to g 2 in perturbation theory. Though we kept the SUŽ n.-models in mind, the computation is identical for any group and representation, and the statement that the OPE closes on the currents themselÕes up to 1-loop order is valid in all ŽM.CGN models. Therefore in the models under consideration, the conservation of the non-local charge is proved whenever among the a i ’s there is no integer greater than one. As a warm-up we repeat the result in the SUŽ n. theories with the fermions being in the fundamental representation. The decomposition of the direct product is:
Ž R m R . m Ž R m R . s Ž1 m R a d j . [ Ž R a d j m 1. [ Ž R a d j m R a d j . [ Ž 1 m 1. .
Ž 15 .
T. Hauerr Physics Letters B 417 (1998) 297–302
301
Table 1 C1 m C 2
1 m 52
26 m 26
0
1 3
a
26 m 273
52 m 52
52 m 324
273 m 273
273 m 324
324 m 324
1
13 9
5 3
16 9
17 9
1
The corresponding Casimirs are C1 s 0;C a d j s 1, which yields a 1, a d j s 0 and a a d j, a d j s 1. The largest integer is 1, this was why we performed the one-loop computation in w4x and found that the quantum charge is conserved in the multiflavor SUŽ n.-models. ŽIn w4x other arguments, like C-parity were also used to rule out quadrilinear operators, which is not needed here since their a are smaller than the largest allowed one.. Now consider the SO Ž n. models with the fermions being in the vector representation. As we expect, here we face more operators than in the SUŽ n.-case. We repeat the decomposition for SO Ž n.:
Ž R m R . m Ž R m R . s Ž1 m R a d j . [ Ž R a d j m 1. [Ž R ad j m R ad j . [ Ž RS m R ad j . [ Ž R ad j m RS . [Ž RS m RS . [ . . .
Ž 16.
In this case R s R and R S stands for the symmetric tensor representation, and we did not list the representations not containing the adjoint here. Furthermore the Casimirs, C1 s 0;C a d j s n y 2;CS s n give the following a ’s:
a 1, a d j s 0, aS , a d j s
a a d j, a d j s 1, n
ny2
,
a S ,S s
nq2 ny2
.
all the a ’s are non-integers except a 1, thus this model also possesses Yangian symmetry. Let us now see how the procedure works for other groups taking F4 first as an example. Let the fermions be in representation 26 which is self-conjugate and the bilinears decompose according to 26 m 26 s 1 [ 26 [ 52 [ 273 [ 324. The quadratic Casimirs are 0,12,18,24,26, respectively and the corresponding a ’s are summarized in Table 1. We can see that the biggest integer is 1 from which one concludes the absence of extra operators in the OPE and the conservation of the non-local charge in the F4-model. It is straightforward to repeat the argument for other groups and show that the same conclusion can be drawn for the G s E7 , R s 56-model however in case of G s E6 , R s 27 and G s E8 , R s 248 we obtain a 650m650 s 2 and a 30380m30380 s 3, respectively, that is, a two and three-loop perturbative calculation is needed in these models. It is clear that one can go on and play around with other groups and representations using a table of group dimensions and indices without difficulty. Note however that, to prove the non-conserÕation of the non-local charge in a specific model one can not avoid going beyond one-loop order in perturbation theory.
Ž 17 .
The largest integer is 5, 3 and 2 for n s 3, 4 and 6, respectively and 1 in all other cases, which proves the vanishing of the OPE-coefficient of extra operators for almost every n; moreover under the mildest assumption about the continuity of the crucial coefficient in terms of n one conjectures its vanishing for all n. This proves the conjecture that the SO Ž n. CGN-models also possess the Yangian algebra w13x Žwhich can be used to prove their integrability. and this is equally true for the multiflavor models as well. We can also consider other representations: as an example take the exotic SO Ž7. model with the fields in the 8 spinor representation. One finds that
6. Conclusion Once a theory possesses ‘‘non-local’’ charges besides the usual ‘‘local’’ ones, it can be shown that, a Yangian symmetry algebra is generated. Due to the non-local nature of the generators however, this is not a direct multiplier of the two-dimensional Poincare´ group w12,13x . This fact reduces the dynamical question about the mass spectrum of a field theory to one on the classification of representations of the underlying algebra Žsee e.g. w14,15x.; and after
302
T. Hauerr Physics Letters B 417 (1998) 297–302
identifying the spectrum, the S-matrix is also highly constrained being proportional to the R-matrix in the given representation. Another approach to the determination of the S-matrix, Žwhich was originally de. uses the fact that the action of veloped by Luscher ¨ the non-local charge on asymptotic one-particle states is straightforwardly computed, in terms of which all the asymptotic matrix elements can be obtained w12,16x. These results lead to the absence of particle production and factorization w3,17x. In this letter, various chiral Gross-Neveu models – which are identified by their symmetry group, representation and multiplicity of the fields – have been considered, and the existence of the non-local generator of the expected Yangian symmetry algebra was discussed. It has long been known that the SUŽ n. one-flavor model possess the Yangian symmetry but one expects that there may be other theories with this property, too. In a previous article we proved this for the multiflavor SUŽ n. model and the method developed there was extended to the whole family in the present paper. The intricate question related to the non-perturbative definition of the model could be answered using finite order perturbation theory Žthanks to asymptotic freedom. – often a one-loop calculation is sufficient. Examples for the latter case were presented proving among others that in the SO Ž n. models the non-local charge can also be defined and that, more exotic theories possess it as well; moreover that, this property is independent of the number of flavors. We also have found CGNmodels where – within this framework – only a higher-loop calculation could decide which class they belong to; and it is an open question whether there
exist generalized CGN quantum theories without the non-local charge generating Yangian symmetry at all.
Acknowledgements I am grateful to J. Balog and P. Forgacs ´ for helpful discussions. This work was partly supported by the Hungarian National Science Fund OTKA, grant No. T19917.
References w1x H.J. de Vega, H. Eichenherr, J.M. Maillet Comm. Math. Phys. 92 Ž1984. 507. w2x H.J. de Vega, H. Eichenherr, J.M. Maillet Nucl. Phys. B 240 Ž1984. 377. w3x M. Luscher, Nucl. Phys. B 135 Ž1978. 1. ¨ w4x T. Hauer, hep-thr9702016, to appear in Nucl. Phys B w5x E. Abdalla, M.C.B. Abdalla, K.D. Rothe Nonperturbative methods in two-dimensional quantum field theory Singapore, World Scientific, 1991. w6x E. Abdalla et al., Nucl. Phys. B 210 Ž1982. 181. w7x E. Abdalla et al., Phys Rev. D 23 Ž1981. 1800. w8x E. Abdalla et al., Phys Rev. D 27 Ž1983. 825. w9x D.J. Gross, A. Neveu, Phys. Rev. D 10 Ž1974. 3235. w10x H.J. de Vega, H. Eichenherr, J.M. Maillet Phys. Lett. B 132 Ž1983. 337. w11x T.L. Curtright, C. Zachos Phys. Rev. D 24 Ž1981. 2661. w12x M. Luscher, unpublished notes. ¨ w13x D. Bernard, Comm. Math. Phys. 137 Ž1991. 191. w14x A. Belavin, Phys. Lett. B 283 Ž1992. 67. w15x T. Nakanishi, Nucl. Phys. B 439 Ž1995. 441. w16x D. Buchholz, J.T. Lopuszanski, Lett. Math. Phys 3 Ž1979. 175. w17x E. Abdalla, A. Lima-Santos, Rev. Bras. Fis 12 Ž1982. 293.